OpenAI releases 372 groups of math results from unreleased frontier AI model

Reactions have been mixed, with excitement over AI’s math skills and questions about how the results should be checked and credited.(Image: Reuters)

After a series of advances in mathematical research, OpenAI on Tuesday, October 6, announced that it was releasing a broad range of new mathematical results produced by one of its internal frontier models. The announcement comes weeks after the company claimed that its model had resolved the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems.

The Sam Altman-led AI startup has released 722 manuscripts spanning 372 groups of results, which the company said include solutions to, or significant progress on, hundreds of open problems in mathematics and theoretical computer science. The results have been published in a GitHub repository with protocols for paper revisions and citations. 

The company also said that to improve how it shares results with the broader math community, it has been consulting an independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study. The group is helping the company develop best practices for releasing AI-generated mathematical research.

OpenAI’s earlier claim about the Navier–Stokes problem sparked both excitement and criticism from mathematicians, particularly over how AI-generated mathematical results should be verified and credited.

What does it mean?

The release highlights the growing ability of AI systems to tackle mathematical research. OpenAI said the internal frontier model was given about 4,000 mathematical problems, with the average result requiring computing equivalent to around three hours of ChatGPT Pro thinking. The company has also published 10 summaries showing how the model approached selected problems.

Many of the proofs have been formalised using Lean, a programming language that allows mathematical proofs to be checked by computers. OpenAI said it will continue adding more formalised proofs to the repository.

According to Scientific American, nearly all of the newly released results were produced by a single AI agent responding to a single prompt. This differs from OpenAI’s earlier Navier–Stokes effort, which involved a large swarm of AI agents and reportedly millions of dollars in computing resources.

Story continues below this ad

The scale of the release has also raised questions about how AI-generated mathematics should be verified and credited. Researchers have called for greater transparency, independent verification and proper acknowledgement of previous human work.

OpenAI has consulted an independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study on these issues. The company said its latest release was informed by the group’s recommendations and includes protocols for revising and citing the papers.

The company also said it plans to continue testing its most advanced models on mathematics and other scientific problems, arguing that AI could eventually give researchers new tools for making discoveries.